Did the Buddha speak Pali? Are the suttas his word verbatim?

The most important and genuine sayings are preserved, albeit scattered around, in Pali Tripitaka. Most are latter commentaries though.

According to renowned Tripitaka translator and scholar Bhikkhu Bodhi (from here):

Scholars regard this language (Pali) as a hybrid showing features of several Prakrit dialects used around the third century BCE, subjected to a partial process of Sanskritization. While the language is not identical to what Buddha himself would have spoken, it belongs to the same broad language family as those he might have used and originates from the same conceptual matrix. This language thus reflects the thought-world that the Buddha inherited from the wider Indian culture into which he was born, so that its words capture the subtle nuances of that thought-world.

Please also see the book "The Authenticity of the Early Buddhist Texts" by Bhikkhu Sujato and Bhikkhu Brahmali, as it contains a detailed analysis on how authentic are the Pali suttas. It may not be verbatim, but it is mostly authentic, according to their analysis.

The abstract states:

This work articulates and defends a single thesis: that the Early Buddhist Texts originated in the lifetime of the Buddha or a little later, because they were, in the main, spoken by the Buddha and his contemporary disciples. This is the most simple, natural, and reasonable explanation for the evidence.

Our argument covers two main areas:
1. The grounds for distinguishing the Early Buddhist Texts (EBTs) from later Buddhist literature;
2. The evidence that the EBTs stem from close to the Buddha’s lifetime, and that they were generally spoken by the historical Buddha.

Most academic scholars of Early Buddhism cautiously affirm that it is possible that the EBTs contain some authentic sayings of the Buddha. We contend that this drastically understates the evidence. A sympathetic assessment of relevant evidence shows that it is very likely that the bulk of the sayings in the EBTs that are attributed to the Buddha were actually spoken by him. It is very unlikely that most of these sayings are inauthentic.

The section called thesis states:

1. That most of the Early Buddhist Texts (EBTs) are authentic.
2. That the EBTs were edited and arranged over a few centuries following the Buddha’s demise. The texts as we have them now are not a verbatim record of the Buddha’s utterances, but the changes are in almost all cases details of editing and arrangement, not of doctrine or substance.
3. That the inauthentic portions of these texts are generally identifiable.
4. That the above points are supported by a substantial and varied body of empirical evidence.
5. That the denial of authenticity is a product of excessive and unreasonable scepticism, not evidence.

1. According to the Buddha's advice, early Buddhists composed and taught the Suttas in their respective vernaculars. So there was not an official language of Buddhism. Wherever Buddhism went, it adopted the local language of that region. Now, Pali was the native language of the region around Vidisha in central India as the Girnar inscription shows. The mother of Ashoka's son Mahendra was form Vidisha. Also, Mahendra started his missionary journey to Sri Lanka from here to reach the West Coast and sail to Sri Lanka. At that time, there was no written canon. So Mahendra just took some monks with him who knew different parts of the canon. Since these monks were from Vidisha region, they learned the Suttas in Pali and they taught these Suttas to the Sri Lankans who mistakenly thought that Pali was the mother tongue of the Buddha. Since then, this belief has been common in the Theravada countries. But it is clear from the above that there is no basis in this belief. The Buddha never went to Vidisha in his lifetime. He was a native of Koshala. So it is likely that he spoke the Koshalan dialect of Prakrit.

2. Pali canon is the canon of the Theravada school which was just one of the many (the traditional number is eighteen but this is clearly an approximate one, the actual number may vary slightly) schools of sectarian Buddhism. These schools started to originate probably one and a half century after the demise of the Buddha. This schism process lasted till the birth of Christ and even afterwards. At present, we do not have the canon of the non-sectarian Buddhism. So we cannot expect whole of the Pali canon as verbatim words of the Buddha. Even so far as Theravada is concerned, the Pali canon was written in the last half of first century BCE, long after the formation of Theravada. Before that it was transmitted orally and might have been changed.

1. The Buddha likely spoke in several dialects that were quite similar to Pali. He was immensely well traveled for his time, so in his travels he likely adopted the native dialect to be better understood by the locals. His words were only a small part of his teaching - his enlightened presence and conduct were likely better vessels of his teaching.

2. The Pali canon is a consolidation of his talks, created only after his death. The content is very likely the same, but the rendition is only as Ananda and other disciples remembered him. The standardized repetitive form was created as an aid to rote memorization and oral rendition. There was certainly doubt in their minds as they set down the teachings that they'd not be able to replicate the vastness of the Buddha's being in mere words, but all the same they resolved to make a sincere effort, and the Canon that emerged was a result of seven months of deliberation.

See https://en.wikipedia.org/wiki/First_Buddhist_council

Pali (Pāli) is a Prakrit native to the Indian subcontinent. It is widely studied because it is the language of many of the earliest extant literature of Buddhism as collected in the Pāli Canon or Tipiṭaka and is the sacred language of Theravada Buddhism.

Many Theravada sources refer to the Pāli language as "Magadhan" or the "language of Magadha". This identification first appears in the commentaries, and may have been an attempt by Buddhists to associate themselves more closely with the Maurya Empire. The Buddha taught in Magadha, but the four most important places in his life are all outside of it. It is likely that he taught in several closely related dialects of Middle Indo-Aryan, which had a high degree of mutual intelligibility. There is no attested dialect of Middle Indo-Aryan with all the features of Pāli. Pāli has some commonalities with both the western Ashokan Edicts at Girnar in Saurashtra, and the Central-Western Prakrit found in the eastern Hathigumpha inscription. 

Whatever the relationship of the Buddha's speech to Pāli, the Canon was eventually transcribed and preserved entirely in it, while the commentarial tradition that accompanied it (according to the information provided byBuddhaghosa) was translated into Sinhaleseand preserved in local languages for several generations. In Sri Lanka, Pāli is thought to have entered into a period of decline ending around the 4th or 5th century (as Sanskrit rose in prominence, and simultaneously, as Buddhism's adherents became a smaller portion of the subcontinent), but ultimately survived. The work of Buddhaghosa was largely responsible for its reemergence as an important scholarly language in Buddhist thought. The Visuddhimagga, and the other commentaries that Buddhaghosa compiled, codified and condensed the Sinhalesecommentarial tradition that had been preserved and expanded in Sri Lanka since the 3rd century BCE.

Source: https://en.wikipedia.org/wiki/Pali

See also, http://www.vridhamma.org/Pali-Overview-94-Seminar